Erosion of Deccan Traps determined by river geochemistry: impact on the global climate and the 87 Sr/ 86 Sr ratio of seawater

Transcription

1 Earth and Planetary Science Letters 188 (2001) 459^474 Erosion of Deccan Traps determined by river geochemistry: impact on the global climate and the 87 Sr/ 86 Sr ratio of seawater Cëline Dessert a; *, Bernard Duprë a, Louis M. Franc ois a;b, Jacques Schott a, Jëroªme Gaillardet c, Govind Chakrapani a;d, Sujit Bajpai e e a LMTG, Universitë Paul Sabatier, 38 rue des 36 Ponts, Toulouse, France b LPAP, Universitë de Lie ge, 5 avenue de Cointe, 4000 Lie ge, Belgium c Laboratoire de Gëochimie et Cosmochimie, IPGP, 4 place Jussieu, Paris, France d Department of Earth Sciences, University of Roorkee, Roorkee , India School of Environmental Sciences, Jawaharlal Nehru University, New Delhi , India Received 21 November 2000; accepted 19 March 2001 Abstract The impact of the Deccan Traps on chemical weathering and atmospheric CO 2 consumption on Earth is evaluated based on the study of major elements, strontium and 87 Sr/ 86 Sr isotopic ratios of the main rivers flowing through the traps, using a numerical model which describes the coupled evolution of the chemical cycles of carbon, alkalinity and strontium and allows one to compute the variations in atmospheric pco 2, mean global temperature and the 87 Sr/ 86 Sr isotopic ratio of seawater, in response to Deccan trap emplacement. The results suggest that the rate of chemical weathering of Deccan Traps (21^63 t/km 2 /yr) and associated atmospheric CO 2 consumption (0.58^2.54U10 6 mol C/km 2 /yr) are relatively high compared to those linked to other basaltic regions. Our results on the Deccan and available data from other basaltic regions show that runoff and temperature are the two main parameters which control the rate of CO 2 consumption during weathering of basalts, according to the relationship: f ˆ R f UC 0 exp 3Ea R 1 T where f is the specific CO 2 consumption rate (mol/km 2 /yr), R f is runoff (mm/yr), C 0 is a constant ( = 1764 Wmol/l), Ea represents an apparent activation energy for basalt weathering (with a value of 42.3 kj/mol determined in the present study), R is the gas constant and T is the absolute temperature (³K). Modelling results show that emplacement and weathering of Deccan Traps basalts played an important role in the geochemical cycles of carbon and strontium. In particular, the traps led to a change in weathering rate of both carbonates and silicates, in carbonate deposition on seafloor, in Sr isotopic composition of the riverine flux and hence a change in marine Sr isotopic composition. As a result, Deccan Traps emplacement was responsible for a strong increase of atmospheric pco 2 by 1050 ppmv followed by a new steady-state pco 2 lower than that in pre-deccan times by 57 ppmv, implying that pre-industrial atmospheric pco 2 would have been 20% higher in the absence of Deccan basalts. pco 2 evolution was accompanied by a rapid * Corresponding author. Fax: ; dessert@lmtg.ups-tlse.fr X / 01 / $ ^ see front matter ß 2001 Elsevier Science B.V. All rights reserved. PII: S X(01)00317-X

2 460 C. Dessert et al. / Earth and Planetary Science Letters 188 (2001) 459^474 warming of 4³C, followed after 1 Myr by a global cooling of 0.55³C. During the warming phase, continental silicate weathering is increased globally. Since weathering of continental silicate rocks provides radiogenic Sr to the ocean, the model predicts a peak in the 87 Sr/ 86 Sr ratio of seawater following the Deccan Traps emplacement. The amplitude and duration of this spike in the Sr isotopic signal are comparable to those observed at the Cretaceous^Tertiary boundary. The results of this study demonstrate the important control exerted by the emplacement and weathering of large basaltic provinces on the geochemical and climatic changes on Earth. ß 2001 Elsevier Science B.V. All rights reserved. Keywords: erosion; rivers; geochemistry; Sr-87/Sr-86; Deccan Traps; Global change; weathering; paleoclimatology; K-T boundary 1. Introduction Continental erosion is a major geological process which is responsible for landscape evolution and exerts a major control on the transport of erosion matters from continents to the oceans and on the cycles of many chemical elements, including carbon, at the Earth's surface. Moreover, as chemical weathering involves consumption of atmospheric CO 2, a greenhouse gas, it probably strongly in uences the secular evolution of the Earth's climate. Over geologic time scales, three major processes consume atmospheric CO 2 : (1) the chemical weathering of continental silicate rocks [1^3], (2) the organic carbon burial during sediment deposition (e.g. [4^6]) and (3) the chemical weathering of the oceanic crust [7,8]. Because climatic conditions are dependent on the atmospheric CO 2 levels, several authors have tried to model the evolution of the carbon cycle and other elements over geologic time [2,8^10]. In order to quantify the role of silicate weathering in the carbon cycle, it is necessary to study the geochemistry of rivers and to determine the laws governing chemical weathering. Many studies have focused on river geochemistry on a global [1,11^ 18] and smaller scale [19^23]. These studies suggest lithology to be a dominant parameter which controls intensity of erosion and atmospheric CO 2 consumption, followed by climatic parameters (runo and temperature). The importance of lithology is emphasised in the studies of Meybeck [19] and Amiotte-Suchet and Probst [24] wherein it is pointed out that granite and gneiss are less easily erodible and hence consume less CO 2 compared to the volcanic rocks. In recent years, weathering of basaltic rocks has been addressed in several studies, in particular those of Louvat [21,25] on four volcanic islands: Iceland, Java, la Rëunion and Sao Miguel, and that of Gislason et al. [20] on Iceland. These authors have shown that basaltic weathering is a major CO 2 sink on time scales of several million years because basaltic rocks weather easily. Furthermore, Taylor and Lasaga [26] have shown, from the modelling of the Columbia River basalt, that weathering of large igneous provinces can play a signi cant role in the evolution of the marine Sr isotope record. The aim of this study was to quantify the in uence of the weathering of the Deccan Traps on the chemistry of seawater and the atmospheric CO 2 mass balance. With this purpose, we studied river geochemistry in the Deccan Traps, one of the largest continental ood basalts on Earth, erupted around 65.5 Myr ago close to the time of the Cretaceous^Tertiary boundary (KT boundary [27,28]). The present-day rates of erosion and atmospheric CO 2 consumption are approximated from the chemistry and the Sr isotope composition of the main rivers draining the Deccan Traps. These results are used in a model presented in this study to calculate the evolution of atmospheric CO 2, global surface temperature, carbonate deposition on sea oor and Sr isotopic composition of seawater following Deccan Traps emplacement. To validate the model, the simulation of oceanic 87 Sr/ 86 Sr evolution is compared with the actual marine Sr isotope record around the KT boundary [29^31]. 2. General setting of Deccan Traps The Deccan volcanic province is located in the

3 C. Dessert et al. / Earth and Planetary Science Letters 188 (2001) 459^ Fig. 1. Location of sampling points in the Deccan Traps volcanic province. western and central parts of India (Fig. 1). These traps form a high plateau with an average elevation of 750 m. Presently, western India is exposed to a monsoonal climate. The most humid period, between June and September, witnesses the major part of the annual rainfall, with a maximum in July (1150 mm). The temperature is relatively high with an annual mean of 25³C and a maximum in May (35³C). K^Ar, 40 Ar^39 Ar and recent Re^Os geochronology coupled with palaeomagnetic and palaeontological studies suggest that the Deccan Traps were erupted around the KT boundary 65.5 Myr ago [27,28,32], when the Indian plate was moving northwards [33]. This volcanic event was apparently quite short on a geological time scale, spanning less than 1 Myr. The Deccan Traps are one of the largest basaltic provinces on the Earth's surface. They have a present-day volume of V10 6 km 3 and cover an area of V5U10 5 km 2. Courtillot et al. [27] suggested that the total initial volume of lava may have reached V3U10 6 km 3. Therefore, some two thirds of the initial basalt must have disappeared in the last 65 Myr. The thickness of the lava pile varies from 200 m or less in the east to 1500^2000 m in the west [27,34]. Javoy and Michard [35] estimated that the total amount of CO 2 outgassed during eruption was 1.6U10 18 mol, which would represent half of the total ocean content. The western Deccan Traps have been extensively studied in terms of geochemistry and stratigraphy [34,36^38]. Basalts overlie crystalline basement consisting largely of granites and gneisses, and sediments. The basalts are mainly of tholeiitic composition and contain phenocrysts of plagioclase, clinopyroxene, altered olivine, opaque minerals and altered glass. Stratigraphy is based on the chemical and/or isotopic composition of the ows. The basaltic sequence is divided into ve sub-horizontal distinct formations: from base to top, the Bushe, Poladpur, Ambenali, Mahabaleshwar and Panhala. The lowermost formations have high Sr isotopic ratios, with all Bushe formations having 87 Sr/ 86 Sr values higher than and all Poladpur formations having 87 Sr/ 86 Sr values in the range 0.705^ The other three formations have 87 Sr/ 86 Sr values around [34,36]. Older sequences outcrop in the north, with a progressive overstepping, and younger towards the south [34,38]. Cox and Hawkesworth [36] have shown that observed chemical variations are due to heterogeneity of mantle sources, variations in the degree of crustal contamination and e ects of fractional crystallisation. These formations are separated in some instances

4 462 C. Dessert et al. / Earth and Planetary Science Letters 188 (2001) 459^474 by lateritic soil horizon and regional lateritic carapaces are preserved in the coastal region [39]. 3. Field and laboratory techniques 3.1. Sampling Water and suspended and bottom sediments were collected during July 1998 in three signi cant watersheds, namely the Godavari, Narmada and Tapti (Fig. 1). The Godavari drainage area is dominated by basaltic rocks associated with crystalline rocks of the Indian Shield, `Gondwana' sediments and Proterozoic sediments. The main tributaries are the Wardha (referring to no. 19 in Fig. 1) and the Pench (no. 25). The Wainganga (no. 20) and the Kanhan (no. 21) do not ow through any basaltic lithology and the Ramakona (no. 23) partly drains basalts. The Narmada was sampled twice within an interval of 10 days (no. 2 and no. 28). All Narmada sampled tributaries drain basaltic rocks except the Silverfall (no. 26) and the Denwa (no. 27). The Tapti (no. 8 and no. 12) drains only Deccan Traps and the Purna (no. 9 and no. 18) is its main tributary Analyses ph and alkalinity were measured in the eld during sampling. The alkalinity was determined by acid^base titration (Gran method). Water samples were ltered on site through 0.2 Wm cellulose acetate lters (142 mm diameter) using a Sartorius ltration unit made of Te on. Filtered samples were acidi ed to ph 2 with HNO 3 and stored in acid-washed polypropylene bottles. The samples for anion determination were not acidi ed. Anions and cations were measured by ion chromatography with a precision better than 5%. The accuracy of the analysis was assessed by running the SLRS-3 and the SLRS-4 riverine standards. Dissolved silica concentrations were determined by spectrophotometric measurement. Trace elements were measured by ICP-MS after addition of an indium standard solution. 87 Sr/ 86 Sr isotopic ratios were determined using mass spectrometry. 4. Results Concentrations of major elements and Sr isotopic compositions are listed in Table 1. The river waters are mostly alkaline with ph ranging from 7.54 to Waters sampled in rivers owing through the basaltic terrains have high ionic charge (4 V1839^6312 Weq/l) and those draining partly through sediments and crystalline rocks have lower ionic charge (153^2722 Weq/l). Waters are characterised by their speci c charge balance, as indicated by the normalised inorganic charge balance (NICB = ( )/4 ). Values are generally close to 5%, except for the Ramakona (no. 23) and Kulbakera (no. 24), where the imbalance can be attributed to less precise alkalinity measurements in the eld. In these cases, we recalculated the alkalinity from the charge balance. The dissolved silica concentration varies between 187 and 836 Wmol/l for the rivers draining basalts, and between 100 and 314 for other rivers. Concentrations in rivers owing through the Deccan Traps are similar to those for Rëunion Island rivers (200^800 Wmol/l [21]). HCO 3 3 (1148^5494 Wmol/l) is always the dominant anion species, followed by Cl 3 (53^1568 Wmol/l), SO 23 4 (10^268 Wmol/l), NO 3 3 (4^141 Wmol/l), and F 3 (1.8^21 Wmol/l). The lower anionic concentration characterises rivers draining sediments and crystalline rocks. This observation is also true for cationic content. Among major cations, Na is dominant (47^2794 Wmol/l). Ca 2 (31^919 Wmol/l) and Mg 2 (17^1271 Wmol/l) have the same range of concentrations whereas concentrations of K range between 7 and 163 Wmol/l. The sum of Na and K represents around 28% of the total cation charge. The cation concentrations are lower than those reported in a previous study on the Narmada and Tapti rivers (e.g. Na V450^7600 Wmol/l [40]). This di erence can be explained by the fact that sampling for the previous study was performed in May, when river discharge is lower compared to that during the south west monsoon period. The concentrations of Sr vary between 1.13 and 3.51 Wmol/l for rivers draining basaltic lithology and between 0.07 and 1.47 for the other rivers.

5 C. Dessert et al. / Earth and Planetary Science Letters 188 (2001) 459^ Table 1 Chemical composition of dissolved load Sample and location ph Na NH4 K Mg Ca F Cl NO3 SO4 HCO3 SiO NICB Sr Sr/ 86 Sr Na* K* Mg* Ca* SO 4* Wmol/l Wmol/l Wmol/l Wmol/l Wmol/l Wmol/l Wmol/l Wmol/l Wmol/l Wmol/l Wmol/l Weq/l Weq/l % Wmol/l Wmol/l Wmol/l Wmol/l Wmol/l Wmol/l Kolar (1) nd Narmada (2) nd þ Ganjal (3) nd þ Machak (4) nd nd þ Kalimachak (5) nd Khagni (6) Chotatawa (7) nd nd þ Tapti (8) þ Purna (9) Bori (10) nd þ Panjkra (11) þ Tapti (12) nd Aranawati (13) nd Aner (14) þ Wagurr (15) nd Nalganga (16) Mun (17) þ Purna (18) þ Wardha (19) nd þ Wainganga (20) nd Kanhan (21) nd Jam (22) nd þ Ramakona (23) nd Kulbakera (24) nd þ Pench (25) nd þ Silverfall (26) nm nd nm Denwa (27) nm nm Narmada (28) nd þ * Indicates concentration corrected from the rain input; nd indicates concentration below detection limit; nm indicates concentrations not measured.

6 464 C. Dessert et al. / Earth and Planetary Science Letters 188 (2001) 459^474 The rivers of the Deccan Traps have higher Sr concentrations than those of the basaltic Rëunion, Sao Miguel and Iceland islands (0.026^0.5 Wmol/l, 0.18^0.79 Wmol/l and 0.015^0.85 Wmol/l respectively [25]). However, the Sr concentrations measured in this study are similar to those reported for Java island rivers (0.63^4.08 Wmol/l [25]). The 87 Sr/ 86 Sr isotopic ratios of the rivers draining Deccan Traps vary between and Chemical weathering rates and isotopic chemistry of Sr 5.1. Chemical weathering rates To quantify the chemical weathering rates, the chemical compositions of river are corrected for atmospheric inputs. This correction was carried out using Cl concentrations and oceanic (X)/(Cl) ratio (where X = Na, Mg, SO 4, Ca or K) and assuming that all the Cl content of the dissolved load has an atmospheric origin. This hypothesis is validated by the very low Cl concentrations of basaltic rocks [34,36]. Cl concentrations in the Kalimachak and Panjkra rivers are very high ( s 1100 Wmol/l, Table 1), suggesting an anthropogenic origin for chloride. Therefore, these rivers were not considered in this study. The corrected concentrations of Na, Mg, SO 4, Ca and K are listed in Table 1. The total dissolved solids (TDS w ) have been calculated from the concentrations of the major dissolved elements (Na, K, Ca, Mg, SiO 2 and SO 4 ) originating from the weathering of Deccan basalts. TDS w values range from 46 to 136 mg/l, with a mean value of 81 mg/l (Table 2). From the data of Dekov et al. [41], we have estimated a mean annual runo value (net ow of water out of watershed) of 463 mm/yr for the Deccan Traps region. This value is similar to that for Narmada (452 mm/yr [40]). From mean runo and TDS w values, we have calculated the speci c chemical weathering rates (rate per unit surface area) listed in Table 2. These rates range between 21 and 63 t/km 2 /yr, with a mean value of 37 t/km 2 /yr. The annual average of chemical erosion ux that Table 2 TDS w values (total dissolved solid after rain input corrections), chemical erosion rates and associated atmospheric CO 2 consumption rates deduced from the dissolved load for the monolithologic rivers draining Deccan basalts Sample and location TDS w Chemical weathering rate CO 2 consumption rate mg/l t/km 2 /yr 10 6 mol/km 2 /yr Kolar (1) Ganjal (3) Machak (4) Khagni (6) Chotatawa (7) Tapti (8) Purna (9) Bori (10) Tapti (12) Aranawati (13) Aner (14) Wagurr (15) Nalganga (16) Mun (17) Purna (18) Wardha (19) Jam (22) Kulbakera (24) Pench (25) Mean

7 C. Dessert et al. / Earth and Planetary Science Letters 188 (2001) 459^ results from the weathering of the Deccan area is close to 18.5U10 6 t/yr (by considering an area of 5U10 5 km 2 for these basalts). The speci c chemical erosion rates for Deccan Traps are higher than those for most of the world's largest rivers (25 t/km 2 /yr for the Amazon and 6 t/km 2 /yr for the Zaire [3]) or similar to those for several mountainous rivers (42 and 46 t/km 2 /yr respectively for the Ganges and the Brahmaputra [3]). If Deccan rivers are compared with other small rivers draining basalts, it appears that chemical weathering rates for Deccan are similar to those for Sao Miguel (35 t/km 2 /yr [25]) and Iceland rivers (41 and 55 t/km 2 /yr [20,25]). Rivers of Rëunion and Java islands exhibit higher chemical weathering rates (102 and 326 t/km 2 /yr [21,25]) because of their higher runo, temperature and relief. Thus, based on the present study and on studies of other basaltic rivers, we conclude that the weathering of the Deccan Traps and, more generally, of basalts is of great importance to the total ux of chemical elements washed to the ocean Flux of strontium and 87 Sr/ 86 Sr isotopic ratios The speci c ux of strontium derived from the weathering of Deccan Traps ranges between 361 and 1080 mol/km 2 /yr, with an average of 751 mol/ km 2 /yr. The total ux of strontium is equal to 3.75U10 8 mol/yr and corresponds to 3.7% of strontium derived from silicate weathering [3]. It is interesting to note that the Deccan area represents only 0.5% of the total surface of continents. Thus, the ux of strontium originated from the weathering of the Deccan basalts should not be neglected in the global balance of strontium to the ocean. The 87 Sr/ 86 Sr isotopic ratios measured in Deccan rivers vary between and Sr/ 86 Sr in the Kanhan (samples 21^25) is signi cantly higher ( s 0.710) than that in the samples from the Tapti and the Narmada basins, where it varies slightly with many tributaries having 87 Sr/ 86 Sr close to All these values re ect the chemical erosion of the basaltic rocks which have unusually high 87 Sr/ 86 Sr isotopic ratios ranging between and [34,36]. Therefore, the 87 Sr/ 86 Sr ratio of rivers draining Deccan Traps is substantially higher than those of volcanic islands studied by Louvat (0.7035^ [25]). 6. Atmospheric CO 2 consumption by chemical weathering The consumption rate of atmospheric CO 2 associated with the chemical weathering of basalts was calculated from riverine HCO 3 3 concentrations. During the weathering of silicate rocks, all dissolved bicarbonates originate from atmospheric/soil CO 2. Values of CO 2 consumption rates listed in Table 2 are very high, ranging from 0.58 to 2.54U10 6 mol/km 2 /yr with a mean value of 1.26U10 6 mol/km 2 /yr. The annual average CO 2 consumption that results from the weathering of the Deccan Traps basalt province is close to 0.63U10 12 mol/yr. The amount of CO 2 consumed by the weathering of Deccan basalts represents 5% of total CO 2 consumption ux derived from silicate weathering [3]. It must be highlighted that this ux is higher than the silicate alkalinity ux determined for the Amazon (2.75% [3]), the Ganges^Brahmaputra system (2.3% [17]) and the Congo (1.7% [3]). We have plotted in Fig. 2 atmospheric CO 2 consumption rates as a function of runo for the Deccan region and small rivers draining only basaltic or granitic lithology [19^21,23,25,42]. It can be seen that lithology, runo and temperature exert a major control on atmospheric CO 2 consumption. It rst appears that rivers draining basaltic rocks have higher HCO 3 3 concentrations than those draining granitic rocks; HCO 3 3 concentrations of `basaltic rivers' vary between 400 and 2700 Wmol/l, whereas those of `granitic rivers' vary between 40 and 500 Wmol/l. Note also that, for a given runo value, CO 2 consumption rates during basalt weathering are signi cantly higher than those for granites. This rst observation con- rms the important in uence of the basaltic lithology on CO 2 consumption rates as noted previously by several authors [19^21,24]. Fig. 2 also highlights the in uence of runo and temperature on CO 2 consumption rates by basalts. For a given temperature, bicarbonate

8 466 C. Dessert et al. / Earth and Planetary Science Letters 188 (2001) 459^474 Fig. 2. Plots of mean atmospheric CO 2 consumption rates versus runo for rivers draining basaltic or granitic lithologies. a: [25]; b: [42]; c: [19]; d: [20]; e: [23]. The values in parentheses correspond to the main surface temperature of each region. Dashed straight lines reported in this gure represent constant HCO 3 3 concentrations from 50 to 5000 Wmol/l. concentrations are constant and independent of runo. Thus, CO 2 consumption rates appear to be approximately proportional to runo. Unlike runo, HCO 3 3 concentrations are directly dependent on temperature. Indeed, if we consider a constant value of runo, the observed increase of CO 2 consumption rate re ects an increase of temperature. The temperature e ect can be more easily observed in Fig. 3 where, by analogy with experimental studies on the kinetics of mineral weathering, the logarithm of HCO 3 3 concentrations has been plotted versus the reciprocal of absolute temperature (1/T in ³K) for rivers draining basaltic rocks. A good linear relationship is observed between the two parameters. From the slope of this relation, a value of Ea = 42.3 kj/mol can be deduced for the apparent activation energy of basalt weathering. The relationship between HCO 3 3 concentrations and temperature (T) can be de ned as: logc HCO 3 3 ˆ 3 EaUln10 b 1 RT where R represents the gas constant and b the intercept of the linear regression ( = 10.66). The speci c rate of CO 2 consumption of basalts f (mol/km 2 /yr) can be expressed as: f ˆ R f UC HCO where R f (mm/yr) represents the value of runo. Combining Eqs. 1 and 2 gives: f ˆ R f UC 0 exp 3Ea 1 R T where C 0 =10 b3ea=298rln10 = 1764 Wmol/l. This equation is not felt to be valid for the weathering rate on granites. In fact, parameters other than runo and temperature in uence the chemical weathering of granites such as soil thickness and physical erosion intensity [13,15,22]; hence, it is di cult to characterise weathering of granite with a simple relation as obtained for basalt.

9 C. Dessert et al. / Earth and Planetary Science Letters 188 (2001) 459^ log HCO 3 (µmol/l) E ect of Deccan Traps emplacement on climate and chemistry of oceans A simple model is next generated to calculate the evolution of atmospheric CO 2 level, surface temperature and strontium isotopic composition of seawater following Deccan Traps emplacement. Using the results inferred from the dissolved load (Section 5) of Deccan Traps rivers, this modelling work allows us to quantify the impact of trap emplacement on global evolution Model description T ( C) /T x 1000 (K ) Ea = 42.3 kj/mol Fig. 3. Logarithm of HCO 3 3 concentrations (in Wmol/l) versus reciprocal of absolute temperature for rivers draining basaltic rocks [19,20,23,25,42]. From the slope of the linear regression, we estimate an activation energy of 42.3 kj/mol. The model used in this work is a modi ed and adapted version of those developed by Franc ois and Walker [8] and Franc ois et al. [43]. In this work, we model the geochemical cycles of inorganic carbon, alkalinity and strontium. In the case of carbon, the weathering of carbonates (F cw ) is the largest source of inorganic carbon. However, this ux is not the primary source driving the evolution of the system, since departures from equilibrium in the weathering^deposition cycle of carbonates remain limited even in the transient experiments performed here. The second source of inorganic carbon is volcanic CO 2 emission. Let F D vol be the CO 2 emission from Deccan Traps and F R vol the CO 2 emission from the rest of the world, including mid-ocean ridges. In the model reference simulation, we assume that the time span of Deccan Traps emplacement is 10 5 years [32]. The major sink of inorganic carbon is carbonate deposition on the ocean oor (F cd ). This ux is calculated in the model, based on the lysocline depth for calcite and aragonite, in a simple ocean sub-model which is outlined below. In the long term, the carbon budget in a given reservoir is described by a di erential equation of the form: dc T ˆ F D vol dt F R vol F cw3f cd 4 where C T is the amount of carbon in the ocean^ atmosphere system. The major sources of alkalinity are the weathering of continental carbonates and silicates (F cw and F sw ). F D sw is the alkalinity ux from Deccan basalt weathering. As for carbon, the carbonate deposition on the sea oor (F cd ) is the major sink of alkalinity. The long-term budget can be written as: da T ˆ 2 F D sw dt F R sw F cw3f cd 5 where A T is the amount of alkalinity in the ocean. The source of ocean strontium is the weathering of silicate and carbonate rocks and its major sink is the carbonate precipitation. The strontium budget is hence determined by: dsr T ˆ k D sw dt F D sw kr sw F R sw k cwf cw 3k cd F cd 6 where k D sw, kr sw, k cw and k cd are proportionality factors used to transform carbon or alkalinity uxes into strontium uxes [44]. They are considered constant and are obtained from the ratios between strontium and carbon uxes of today (Table 3). The evolution of the 87 Sr/ 86 Sr isotopic ratio of ocean water over geologic time can be written as: Sr T dr oc dt ˆ r D sw 3r oc k D sw F D sw rr sw 3r oc k R sw F R sw r cw 3r oc k cw F cw r sfw 3r oc k sfw F R vol 7

10 468 C. Dessert et al. / Earth and Planetary Science Letters 188 (2001) 459^474 k sfw F R vol is the exchange ux of Sr during sea oor basalt alteration which is considered to be proportional to the non-deccan volcanic CO 2 ux F R vol. r oc is the strontium isotopic ratio of the ocean [45] and r D sw, rr sw, r cw and r sfw are the mean strontium isotopic ratios originating from Deccan basalts, silicate weathering, carbonate weathering and sea- oor basalt alteration [46]. Note that carbonate deposition is not taken into account in the isotopic budget (Eq. 7) because it is assumed that no fractionation occurs during precipitation. The diagenetic ux is small enough that it can be ignored with no signi cant lack of accuracy in Eqs. 6 and 7. The values of some present-day model parameters are given in Table 3 with relevant references. The riverine 87 Sr/ 86 Sr ratio coming from Deccan Traps weathering (r D sw ) used in the model is the present-day ratio. Because of the low ratio Rb/Sr of Deccan basalts [34,36], the chronological correction is insigni cant and we can suppose that the riverine 87 Sr/ 86 Sr ratio coming from Deccan Traps weathering has not signi cantly changed over the last 65 Myr. The oceanic 87 Sr/ 86 Sr ratio before Deccan Traps emplacement was determined from the actual marine Sr isotope record 65 Myr ago [45] and is equal to From this value and the present-day uxes of strontium, we have calculated the pre-deccan 87 Sr/ 86 Sr ratio of global riverine ux (0.7103) and riverine ux coming from silicate weathering excluding Deccan (r R sw = ). Over geologic time, weathering uxes are affected by various factors such as continental surface (A), global average runo per unit area (R f ), surface temperature (T) and atmospheric CO 2 concentration (pco 2 ). In this model, we use the BLAG relations [2] that link R f to T and pco 2 to T: R f R f0 ˆ 1 0:038U T3T 0 T ˆ T 0 2:88Uln pco 2 pco 2;0 8 9 where the 0 subscript refers to present-day values (280 ppmv and 288 K respectively for pre-industrial pco 2 and T). In the case of silicate rocks, weathering uxes are proportional to runo, temperature and continental area. The global ux (F sw ) is the sum of the Deccan Traps contribution and that of the rest of the world. Each contribution is described Table 3 Main parameters of the geochemical model Parameter Value Reference CO 2 consumption ux by carbonate weathering 12.3U10 12 mol/yr [3] CO 2 consumption by silicate weathering 11.7U10 12 mol/yr [3] CO 2 release from Deccan Traps pulses 1.6U10 18 mol [35] CO 2 consumption by weathering of Deccan basalts 0.63U10 12 mol/yr This study Riverine ux of Sr 28.4U10 9 mol/yr Calibrated from CO 2 ux and k values k cw 1/1250 [44] k R sw 1/305 [44] k D sw 1/840 This study k sfw 1/140 [44] Oceanic 87 Sr/ 86 Sr ratio before Deccan Traps emplacement, r oc [45] Global riverine 87 Sr/ 86 Sr ratio before Deccan Traps emplacement Calibrated to obtain an oceanic Sr/ 86 Sr of Riverine 87 Sr/ 86 Sr ratio coming from carbonate weathering, r cw [46] Riverine 87 Sr/ 86 Sr ratio coming from Deccan Traps weathering, r D sw This study Riverine 87 Sr/ 86 Sr ratio coming from silicate weathering, r R sw (excluding Deccan) Calibrated to obtain a riverine Sr/ 86 Sr of Sr/ 86 Sr ratio coming from sea oor basalt weathering, r sfw [45]

11 C. Dessert et al. / Earth and Planetary Science Letters 188 (2001) 459^ by the following equation: F i sw ˆ A U R f Uexp 3 Ea i 1 A 0 R f0 R T 3 1 T 0 F i sw0 10 respective lysocline (i.e. these uxes are proportional to the fraction of the sea oor area located above the aragonite or calcite lysocline), as well as of coral reef formation on the equatorial shelf. with i = D or R. A/A 0 is the area (Deccan Traps or rest of the world) normalised to today's value. R is the gas constant and Ea is the activation energy of the mineral dissolution reaction. An apparent energy of 42.5 kj/mol and 40 kj/mol is used for Deccan basalt dissolution (determined in Section 5) and dissolution of silicates from the rest of the world (unpublished data of Oliva), respectively. Before the emplacement of Deccan basalts, we suppose that this area was constituted by silicates as in the rest of the world (with Ea = 40 kj/mol) and that the system was at steady state. To calculate carbonate weathering, it is assumed that stream water draining carbonate lithologies reaches equilibrium with calcite and an e ective pco 2 value (pco eff ), which is a harmonic mean between soil 2 and atmospheric pco 2. Soil pco 2 is calculated according to Volk's [47] study with a present value 100 times higher than the atmospheric value. The ocean sub-model allows us to estimate atmospheric pco 2 and F cd values for all C T and A T combinations determined at each time step of the numeric resolution (Eqs. 4 and 5). This sub-model contains three surface pools: the atmosphere, the surface ocean and the deep ocean, and distributes instantaneously C T and A T between these three boxes. As these boxes are assumed to be in equilibrium with each other, the model cannot be used to analyse the evolution of the system at time scales shorter than 1000 years (oceanic circulation). Carbonate speciation in each oceanic reservoir is taken into account to calculate atmospheric pco 2 in equilibrium with the surface ocean and to determine the lysocline depth for aragonite and calcite from the calculated CO 23 3 concentrations. Carbon is transferred between the surface and the deep ocean reservoir through advective exchange (ocean circulation) and sedimentation of biological particles. The biological new production is constant in all of the simulations. The ux of carbonate deposition F cd is composed of the aragonite and carbonate tests deposited above their 7.2. Results Reference simulation The reference run combines all the e ects presented previously. The calculated evolutions of surface temperature (³C), atmospheric pco 2 (ppmv), carbonate deposition on sea oor (10 13 mol/yr), surface ocean ph and 87 Sr/ 86 Sr ratio of seawater during the emplacement of the traps are shown in Figs. 4^6. The variation of atmospheric pco 2 is relatively important. In fact, pco 2 increases by 1050 ppmv from its pre-deccan steady-state value. It then progressively decreases to reach a new (post-dec- Fig. 4. The evolution of atmospheric CO 2 partial pressure (ppmv) and surface temperature (³C) calculated from the different model runs described in the text.

12 470 C. Dessert et al. / Earth and Planetary Science Letters 188 (2001) 459^474 Fig. 5. The evolution of the carbonate deposition ux in the ocean (10 13 mol/yr) and surface ocean ph calculated from the di erent model runs. can) steady state after a few million years. Note that only 1.2 Myr is necessary to return to the pre-deccan value. After 1.2 Myr, pco 2 continues to decrease and the new steady-state pco 2 is lower than the pre-deccan one by 57 ppmv, as a result of the very e cient weathering of the Deccan basalts. Thus, the emplacement of Deccan Traps appears to be responsible for a 20% reduction of atmospheric pco 2 (with respect to preindustrial pco 2 ). This decrease is accompanied by a global cooling of 0.55³C. The variation of carbonate deposition is greater than in the case of atmospheric pco 2 and surface temperature (Fig. 5). Indeed, this ux rst decreases by 0.87U10 13 mol/yr (345%) from its pre-deccan value in only years, as a result of a rapid rise of the lysocline. This is followed by a rapid rise in carbonate deposition of 2.05U10 13 mol/yr, resulting in a strong increase of continental erosion. Finally, it progressively decreases to reach steady state after 1.4 Myr. On the other hand, it can be seen that the surface ocean ph decreases rapidly from 8.13 to 7.80 in 0.1 Myr. Note that this decrease is more rapid during the rst years. After 0.1 Myr, the time corresponding to the end of Deccan emplacement, the ph progressively increases to reach a steady-state value of These two parameters demonstrate the immediate response of ocean and continents to regulate the strong input of CO 2 in the atmosphere. After the beginning of Deccan eruption, the 87 Sr/ 86 Sr ratio of seawater increased from its pre-deccan value of to , before coming down to (Fig. 6). This spike is characterised by a duration of some 4 Myr and a peak amplitude of about 7U The transient peak in the Sr signal is related to the strong increase of continental weathering during and after the eruption. As pco 2 decreases, continental weathering rates are reduced leading to a decrease in oceanic 87 Sr/ 86 Sr ratio. Furthermore, as r D sw is lower than r R sw, the weathering of Deccan Traps tends to reduce the global isotopic ratio of strontium originating from silicate weathering. Note that the Deccan Traps cover 0.5U10 6 km 2, but their original area was three or four times larger. In this respect, the impact of these traps is underestimated Test simulations In order to explore the contribution of various parameters, several sensitivity tests were performed such as the duration of the emplacement, the pre-deccan atmospheric pco 2, the contribution of the erosion of continental carbonates and the dependence of silicate weathering on atmospheric pco 2. As most of the results of the simulations are relatively close to that of the reference one, only the two more signi cant ones are presented in this study (Figs. 4^6). The rst important sensitivity test (run 1) was performed to analyse the impact of the duration of emplacement of Deccan Traps. In this test, the 1.6U10 18 mol of CO 2 were released to the atmosphere in 1 Myr instead of 10 5 years in the reference run. The characteristic peak of pco 2 during Deccan Traps emplacement is strongly reduced in

13 C. Dessert et al. / Earth and Planetary Science Letters 188 (2001) 459^ Fig. 6. The evolution of the seawater 87 Sr/ 86 Sr ratio calculated from the di erent model runs. By comparison, the data points represent the marine Sr isotope record taken from Martin and Macdougall [31] around the KT boundary. *: absolute age based on Martin and Macdougall was recalibrated from 65 Ma (KT boundary). this simulation, pco 2 increasing by only 317 ppmv. This smaller variation reduces the impact on global temperature. The return to the post- Deccan steady-state value is longer (1.7 Myr instead of 1.2 Myr) and the spike of carbonate deposition is weak. Indeed, after a small decrease of 5%, carbon deposition ux progressively increases by 0.6U10 13 mol/yr, before returning to steady state, about 2 Myr after the beginning of Deccan eruption. Similarly, in comparison with the reference simulation we do not observe any important variation of surface ocean ph. A delay is also observed in the Sr signal, where the peak appears 1.4 Myr after the beginning of the eruption. This simulation illustrates the role of duration of Deccan Traps emplacement at short time scales. In the second sensitivity test (run 2), the pre- Deccan atmospheric pco 2 is assumed higher than in the reference one. To get high CO 2 levels, the CO 2 release from pre-deccan volcanism (F R vol ) was increased by a factor of 1.5. This larger volcanic ux resulted in a pre-deccan atmospheric CO 2 of 1435 ppmv and a global surface temperature of K. Under such high levels of CO 2, the evolution of pco 2 following Deccan Traps emplacement is perceptibly di erent. Atmospheric pco 2 increases by 1792 ppmv after emplacement and returns to a new steady-state pco 2 lower than that in pre-deccan times by 260 ppmv. On the other hand, the increase of pre-deccan pco 2 has a smaller e ect on the variation of global temperature. Indeed, the warming related to the CO 2 degassing phase is smaller than the reference one (2.3³C instead of 4.1³C) and the global cooling is unchanged. We note that, as a result of higher volcanic pco 2 uxes, the carbonate deposition ux is higher than the reference one. The ph is relatively low (as expected under high pco 2 ) and its variation is smaller. The increase of pre-deccan pco 2 also has an e ect on the

14 472 C. Dessert et al. / Earth and Planetary Science Letters 188 (2001) 459^474 marine Sr record since the spike is characterised by an amplitude (peak at ) and a duration (3 Myr) smaller than the reference one. This simulation illustrates that, even if the increase of pre- Deccan pco 2 has a global e ect on silicate weathering and the Sr cycle, the impact of Deccan Traps emplacement is reduced Comparison between results of model and marine sediment records We have shown that Deccan Traps emplacement had multiple e ects on the global environment. We propose that this event caused a sharp decrease in sedimentation (345% in only 20 kyr), induced by a strong change in lysocline depth. This disturbance and decrease of carbonate deposition has also been observed at the KT boundary by several authors including Smit and Romein [48] and Kaiho et al. [49] who inferred a 13 kyr interval of productivity shutdown. The simulations involve a rapid increase of seawater 87 Sr/ 86 Sr ratio of 8U10 35 with a duration of 4 Myr; therefore the emplacement and weathering of these continental ood basalts may explain the very abrupt KT boundary spike in 87 Sr/ 87 Sr [29^ 31]. In order to validate the model, we compare our simulations of oceanic 87 Sr/ 86 Sr evolution with the actual marine Sr isotope record around the KT boundary of Martin and Macdougall [31] in Fig. 6. To reduce variability due to chronological uncertainty, sediment mixing, analytical uncertainty and diagenesis, the authors have smoothed and replotted the data. In this study the KT boundary is assimilated to the spike in 87 Sr/ 86 Sr at 66.4 Ma. We recalibrated the age of this boundary at 65 Ma, its well known age. The data of Martin and Macdougall show a spike with a peak value of about , a maximum amplitude of order 7U10 35 and a maximum duration of 3 Myr that is relatively similar to the evolution proposed in our model. This study allows us to suggest an origin for the Sr isotope anomaly at the KT boundary. The results of the model show that a number of geochemical characteristics of the KT boundary can be explained by Deccan Traps emplacement. 8. Conclusion This study presents a coupled geochemical investigation of major and trace elements of the main rivers draining the Deccan Traps province as well as a simple model making it possible to quantify the impact of Deccan Traps emplacement. b b b The chemical composition of the dissolved load in the main rivers draining Deccan Traps basalts allows us to calculate chemical weathering rates (21^63 t/km 2 /yr) and associated atmospheric CO 2 consumption rates (0.58^ mol C/km 2 /yr). The results obtained in this study con rm that denudation rates of basaltic rocks are much higher than those for other silicate rocks. Furthermore, it is suggested that runo and temperature are the most important parameters which control the chemical weathering and CO 2 consumption rates, which can be rather accurately described by a simple exponential law: f ˆ R f UC 0 exp 3Ea 1 R T where f is the speci c CO 2 consumption rate (mol/km 2 /yr), R f is runo (mm/yr), C 0 = 1764 Wmol/l, Ea represents an apparent activation energy for basalt weathering (42.3 kj/mol), R is the gas constant and T is the absolute temperature (³K). Our model allows one to quantify the multiple e ects of Deccan Traps emplacement on the global environment. This emplacement is responsible for a 20% reduction of atmospheric pco 2, accompanied by a global cooling of 0.55³C, and has therefore produced a net CO 2 sink on geologic time scales. A peak in the 87 Sr/ 86 Sr isotopic ratio in the oceans following Deccan Traps emplacement is also predicted by the model and indeed allows one to understand the origin of the Sr isotope anomaly at the KT boundary. Overall, this study emphasises the major impact of basalt weathering on geochemical cycles such as the carbon cycle. It shows that the emplace-

15 C. Dessert et al. / Earth and Planetary Science Letters 188 (2001) 459^ ment and the weathering of large basaltic provinces cannot be neglected when attempting to better understand the geochemical and climatic evolution of the Earth. Acknowledgements This work was supported by the French programme PROSE (Programme Recherche Sol Erosion) funded by INSU. We also thank the Foreign O ce and the French Ministry of Education, Research and Technology for funding and providing post-doctoral positions for G.J.C. and S.B. L.M.F. was supported by the FNRS (Belgian National Foundation for Scienti c Research) and obtained from a visiting research position from CNRS. We would like to thank V. Subramanian and the technicians of the School of Environmental Sciences, New Delhi, as well as N. Vigier and T. Allard for assistance in obtaining river samples. We are also grateful to P. Brunet, J. Escalier, P. Oliva, B. Reynier, M. Valladon of the Toulouse laboratory and to F. Capmas, C. Gorge and M. Pierre of IPGP for their support in the di erent analyses. L. Turpin, N. Vigier, J. Viers and P. Oliva are thanked for their helpful comments. We thank V. Courtillot and C. France- Lanord for their very constructive reviews of this manuscript.[ac] References [1] R.M. Garrels, F. Mackenzie, Evolution of Sedimentary Rocks, Norton, New York, [2] R.A. Berner, A.C. Lasaga, R.M. Garrels, The carbonatesilicate geochemical cycle and its e ect on atmospheric carbon dioxide over the past 100 millions years, Am. J. Sci. 284 (1983) 641^683. [3] J. Gaillardet, B. Duprë, P. Louvat, C.J. Alle gre, Global silicate weathering and CO 2 consumption rates deduced from the chemistry of the large rivers, Chem. Geol. 159 (1999) 3^30. [4] C. France-Lanord, L.A. Derry, Organic carbon burial forcing of the carbon cycle from Himalaya erosion, Nature 390 (1997) 65^67. [5] J.M. Hayes, H. Strauss, A.J. Kaufman, The abundance of 13 C in marine organic matter and isotopic fractionation in the global biogeochemical cycle of carbon during the past 800 Ma, Chem. Geol. 161 (1999) 103^125. [6] M.E. Raymo, The Himalayas, organic carbon burial, and climate in the Miocene, Paleoceanography 9 (1994) 399^ 404. [7] A.J. Spivack, H. Staudigel, Low-temperature alteration of the upper oceanic crust and the alkalinity budget of seawater, Chem. Geol. 115 (1994) 239^247. [8] L.M. Franc ois, J.C.G. Walker, Modelling the Phanerozoic carbon cycle and climate: constraints from the 87 Sr/ 86 Sr isotopic ratio of seawater, Am. J. Sci. 292 (1992) 81^ 135. [9] J.C.G. Walker, P.B. Hays, J.F. Kasting, A negative feedback mechanism for the long-term stabilization of Earth's surface temperature, J. Geophys. Res. 86 (1981) 9776^ [10] O. Brevart, C.J. Alle gre, Strontium isotopic ratios in limestone through geological time as a memory of geodynamic regimes, Sci. Geol. Bull. 19 (1977) 1253^1257. [11] R.F. Stallard, J.M. Edmond, Geochemistry of the Amazon 2. The in uence of geology and weathering environment on the dissolved load, J. Geophys. Res. 88 (C14) (1983) 9671^9688. [12] P. Nëgrel, C.J. Alle gre, B. Duprë, E. Lewin, Erosion sources determined by inversion of major and trace element ratios and strontium isotopic ratios in river water: The Congo Basin case, Earth Planet. Sci. Lett. 120 (1993) 59^ 76. [13] J.M. Edmond, M.R. Palmer, C.I. Measures, B. Grant, R.F. Stallard, The uvial geochemistry and denudation rate of the Guyana Shield in Venezuela, Colombia, and Brazil, Geochim. Cosmochim. Acta 59 (1995) 3301^3325. [14] B. Duprë, J. Gaillardet, D. Rousseau, C.J. Alle gre, Major and trace elements of river borne material: The Congo Basin, Geochim. Cosmochim. Acta 60 (1996) 1301^1321. [15] J. Gaillardet, B. Duprë, C.J. Alle gre, A global geochemical mass budget applied to the Congo Basin rivers: erosion rates and continental crust composition, Geochim. Cosmochim. Acta 59 (1995) 3469^3485. [16] J. Gaillardet, B. Duprë, C.J. Alle gre, P. Nëgrel, Chemical and physical denudation in the Amazone River Basin, Chem. Geol. 142 (1997) 141^173. [17] A. Galy, C. France-Lanord, Weathering processes in the Ganges-Brahmaputra basin and the riverine alkalinity budget, Chem. Geol. 159 (1999) 31^60. [18] J.L. Probst, J. Mortatti, Y. Tardy, Carbon river uxes and weathering CO 2 consumption in the Congo and Amazon river basins, Appl. Geochem. 9 (1994) 1^13. [19] M. Meybeck, Composition chimique des ruisseaux non polluës de France, in: Sciences Geologiques 39, Strasbourg, 1986, pp. 3^77. [20] S.R. Gislason, S. Arnorsson, H. Armannsson, Chemical weathering of basalt as deduced from the composition of precipitation, rivers and rocks in SW Iceland, Am. J. Sci. 296 (1996) 837^907. [21] P. Louvat, C.J. Alle gre, Present denudation rates at Rëunion island determined by river geochemistry: basalt